Sucrose metabolism gene families and their biological functions

Sucrose, as the main product of photosynthesis, plays crucial roles in plant development. Although studies on general metabolism pathway were well documented, less information is available on the genome-wide identification of these genes, their expansion and evolutionary history as well as their biological functions. We focused on four sucrose metabolism related gene families including sucrose synthase, sucrose phosphate synthase, sucrose phosphate phosphatase and UDP-glucose pyrophosphorylase. These gene families exhibited different expansion and evolutionary history as their host genomes experienced differentiated rates of the whole genome duplication, tandem and segmental duplication, or mobile element mediated gene gain and loss. They were evolutionarily conserved under purifying selection among species and expression divergence played important roles for gene survival after expansion. However, we have detected recent positive selection during intra-species divergence. Overexpression of 15 sorghum genes in Arabidopsis revealed their roles in biomass accumulation, flowering time control, seed germination and response to high salinity and sugar stresses. Our studies uncovered the molecular mechanisms of gene expansion and evolution and also provided new insight into the role of positive selection in intra-species divergence. Overexpression data revealed novel biological functions of these genes in flowering time control and seed germination under normal and stress conditions.

encode 4-15 SuSy genes [5][6][7] . However, compared with the SuSy family, less information is available for the genome-wide identification and characterization as well as evolution of genes encoding SPS, SPP and UDPGP. Both SPS and SPP are also encoded by small gene families. SPS is a key enzyme for sucrose synthesis. Castleden et al. (2004) carried out phylogenetic analysis based on all available SPS sequences at that time 8 . Subsequently, genomic and EST sequence data from Arabidopsis, rice and maize were employed to identify the SPS genes. The analysis showed that the Arabidopsis and rice genomes encode a total of 4 and 5 members, respectively and gene duplication might play a role in their expansion [8][9][10] . Relatively, fewer reports have been focused on SPPs. Genes have been identified from several species including rice, maize, wheat, barley, Arabidopsis and tomato 11,12 . These species contain at least two SPP genes [12][13][14] . For example, four and three SPP genes have been identified in Arabidopsis and rice, respectively 11 . For the UDPGP families, previous data showed that only one or two UDPGPs existed in plants 15 . However, studies showed that UDP-sugar pyrophosphorylase also catalyzed the same reaction as the enzyme has broad substrates specificity 16 (Litterer et al. 2006). Besides UDP-glucose, it also catalyzes the reversible formation of UDP-galactose, UDP-glucuronic acid, UDP-l-arabinose and UDP-xylose. In addition, these types of enzymes share the same protein domain structure. Thus, in this study, UDPGP refers to these enzymes with the common Pfam domain (PF01704, http://pfam.xfam.org/) structure. Although more and more whole genome sequencing data are available, no genome-wide identification and characterization have been recently reported for SPS, SPP and UDPGP. In addition, the evolutionary analysis on SuSy, SPS and SPP and UDPGP were mainly focus on the investigation among different genus, families, orders and even kingdoms 3,6,8,12,15,17 and little is known on the recent evolution among species, subspecies or varieties belonging to the same species. SuSy, SPS and SPP could be detected in plants and bacteria. However, UDPGP were detected not only in plants and bacteria but also in animals and fungi. Furthermore, all four types of enzymes have been undergone gene duplication but the mechanism is not yet elucidated.
On the other hand, functional characterization of these four gene families has also been carried out in some plant species by overexpression, knockout and silencing of genes. Overexpression of a SuSy gene increased biomass production and/or sucrose content in switchgrass, tobacco, cotton or poplar [18][19][20][21] . Loss of function of a SuSy gene by knockout mutation exhibited various visible phenotypes [22][23][24] . However, no obvious difference in sucrose synthase activity or growth phenotype was observed when one of two SuSy genes was knocked out in Arabidopsis 25 . Overexpression of SPS genes significantly increased sucrose synthesis in certain tissues and led to the changes in phenotypes including increased biomass [26][27][28] . Anti-sense expression of SPS genes or knockout of these genes by T-DNA/retrotransposon insertion repressed sucrose synthesis and resulted in abnormal development 29,30 . Decreased SPP level in transgenic tobacco by RNAi inhibited photosynthesis and altered carbohydrate partitioning with reduced growth rate. SPP activity was also repressed in transgenic potato tubers (Solanum tuberosum) by RNAi, which affected the hexose-to-sucrose ratio upon cold storage with only minor effects on total soluble carbohydrate accumulation 14 .
Antisense suppression of a UDPGP gene in plants could reduce soluble carbohydrate and starch or sucrose content 31,32 . In rice, inactivation of the UGPase1 gene led to genic male sterility 33,34 . Investigation of T-DNA insertion of a UDPGP gene suggested it should be a novel plant cell death regulator 35 . Data also showed that overexpression of a native or foreign UDPGP gene in various plant species could increase plant height 18 and enhance leaf area and leaf-to-shoot biomass ratio 36 or vegetative growth 37 . Generally, limited data are available for studies on biological functions of these four gene families. For some important crops, for example, sweet sorghum, little is known about their biological functions of these four gene families although sucrose metabolism is an important step for stem sugar accumulation in this plant.
In this study, we first genome-widely identified and characterized all genes encoding SuSy, SPS, SPP and UDPGP in 15 species from algae, moss as well as monocot and dicot plants. We then evaluated their expansion mechanisms and evolutionary history. Subsequently, we investigated their selection forces and expression divergence among different tissues or under various stresses to further annotate their biological functions and retention mechanisms of expanded genes. Finally, we investigated their biological functions by overexpressing 15 sorghum genes in the Arabidopsis genome. Our data show that plant species encode different numbers of SuSy, SPS, SPP and UDPGP genes. These four gene families exhibited different expansion and evolutionary history as their host genomes experienced differentiated rates of the whole genome duplication, tandem and segmental duplication, or mobile element mediated gene gain and loss. All these were evolutionarily conserved under purifying selection among species and expression divergence played important roles for gene survival after expansion. However, we have detected recent positive selection in several sites of the sorghum UDPGP gene Sobic.006G213100, suggesting a role of this gene in intra-species divergence. Overexpression of 15 sorghum genes in Arabidopsis showed the roles in biomass accumulation and revealed novel biological functions in seed germination under normal and high salinity stress.

Results
Plant genomes encode different sizes of the SuSy, SPS, SPP and UDPGP families. To identify and characterize the four family members on a genome-wide level, 15 completely sequenced genomes from monocot, dicot plants as well as algae and spikemoss have been selected for such analyses as listed in Table 1. The Hidden Markov model (HMM) searches followed by subsequent confirmation showed that not all plant genomes encoded the all four gene families (Table 1). Generally, all the 15 selected plant genomes have contained three gene families including SuSy, SPP and UDPGP. However, no SPS gene has been detected in the lower plant species C. reinhardtii. For the SuSy gene family, only one gene was detected in both C. reinhardtii (alga) and S. moellendorffii (moss) and most species encode more than 4 SuSy genes with the highest number (31) of SuSy genes in apple (Table 1). For the SPS gene family, most of plants encode 3-5 members whereas up to 7 SPS genes were identified in maize, soybean or apple (Table 1). For the SPP gene family, only one gene was detected in monkey flower and castor bean and the remaining genomes encode 2-5 genes. For the UGPGP gene family, the plant genomes encode at least 3 members whereas up to 15 members were detected in the soybean genome. Besides the members as listed in Table 1, we have also identified numbers of members encoding partial domains of these protein families. Further analysis showed that these members contained no typical domain structure and they also lacked expression evidence. Thus, these members were removed from our further analysis due to the low feasibility of phylogenetic analyses although the rate of gene duplication might be under-estimated.
Four gene families exhibited different expansion and evolutionary history. To classify the members of these four gene families and to facilitate their functional analysis, we reconstructed the phylogenetic relationship in each family ( Fig. 1A-D, Supplementary Fig. S2 A-D) using their corresponding protein domain sequences as described in the Methods. Previous studies identified 3 groups of SuSy genes, which were based on available SuSy members from limited plant species 4,5 . However, the phylogenetic tree based on SuSy members from 15 species showed that a total of 5 groups could be clustered ( Fig. 1A and Supplementary Fig. S2A). Both groups I and II are the main groups consisting of members from all dicot and monocot plants. The remaining three groups are species-specific and they consist of members from limited species. For the SPS gene family, we identified 3 groups ( Fig. 1B and Supplementary Fig. S2B). Among them, group I is the largest one, consisting of the all members from both dicot and monocot plants. Previous data showed that four groups were identified 8 . However, in the study, groups B and C were actually in the same group 8 ; thus, similar to our result. On the other hand, the SPP family was also clustered into three different groups with the group I as the largest one consisting of all members from angiosperm species (Fig. 1C and Supplementary Fig. S2C). As for the UDPGP family, no genome-wide identification was previously reported. We have identified a total of 110 members from 15 species (Table 1) and they could be clustered into two groups ( Fig. 1D and Supplementary Fig. S2D). In each group, at least one member has been included from all the 15 species.
To evaluate the patterns of expansion and evolutionary history of these four gene families, we broke down the phylogeny tree into ancestral units and estimated the most recent common ancestor (MRCA) among species according to the method described by Shiu et al. (2004) 38 . Due to possible gene losses and pseudogenes, which were excluded in the phylogenetic trees, the MRCA members may be under-estimated. Only one gene with locus name Cre12.g524000 was identified in the green algae species C. reinhardtii and the MRCA among the 15 species was estimated as 1 (Fig. 1E). Since no member was detected in the algae species for the SPS family, no MRCA exit among analyzed 15 species (Fig. 1E) Fig. 2A). The data suggested that a large scale of expansion occurred after the divergence of apple from strawberry. A similar situation was also observed in the maize species. However, for soybean and barrel medic (M. truncatula), the SuSy gene expansion might occur before the  divergence of these two species and the whole-genome duplication might contribute to the expansion ( Fig. 2A). On the other hand, our data also showed that most of species required only small families of SuSy, SPS, SPP and UDPGP with less than 10 members each.

Contributions of duplication and transposition to family size.
To explore the mechanisms of these four gene family expansion, we investigated the contributions of various gene expansion ways to the family size. These mechanisms include tandem, segmental and the whole genome duplication, transposition and retrotransposintion. We first surveyed the contribution of tandem duplication to the gene expansion. We examined the physical positions of the family members on corresponding genome chromosomes. The results showed that tandem duplication contributed to the family expansion in 6 of 15 species (Fig. 2A). It mainly contributed to the expansion of both SuSy and UDPGP gene families. However, in the apple genome, it was the main way for the expansion of all four gene families. More than 50% of SuSy, SPS, SPP and UDPGP genes were expanded by tandem duplication. We then analyzed the contribution of segmental duplication to the family expansion. Our data showed that segmental duplication significantly contributed to the expansion of SuSy, SPS, SPP or UDPGP genes in at least 11 species (Fig. 2A). The contribution rates varied depending on species and their encoded gene families, ie. SuSy, SPS, SPP or UDPGP. In R. communis and M. truncatula, only one and two out of the four gene families was expanded by segmental duplication, respectively. For the remaining species, segmental duplication played roles in the expansion of three or four gene families. In some gene families, majority of their expansion was due to segmental duplication. For example, 100% and 83.5% of SuSy genes were expanded by segmental duplication in poplar and soybean, respectively. Besides tandem and segmental duplication, we also surveyed the contribution of the whole genome duplication to the family expansion. Due to the difficulty in distinguishing segmental duplication from the whole-genome duplication followed by massive loss of duplicate genes 39 , we compared the events of genome doubling (green stars in Fig. 2A) or tripling (blue stars in Fig. 2A) with the family size of SuSy, SPS, SPP and UDPGP. The comparison implied that genome doubling or tripling also contributed to the expansion of these four gene families in some species. For example, Z. mays (maize) underwent one more genome doubling event when compared with S. bicolor (sorghum) and as a result, the maize genome encoded more members of SuSy, SPS and UDPGP ( Fig. 2A).
On the other hand, we have also investigated the contribution of transposable elements (TEs) to the expansion of these four gene families. We have analyzed the roles of LTR-retrotransposon, retrogene, Pack-Mule, hAT, Helitron and CACTA elements in the expansion of the four gene families. Generally, TE-mediated family expansion has been detected in 13 out of 15 species (Fig. 2A). However, only one to four genes were expanded by mobile elements in each species for each gene family. Thus, TE-mediated expansion might not contribute to the large scale of family expansion. Further investigation showed that there exited 4 types of TE-mediated gene expansions as shown in Fig. 2B-D. In type 1, a whole gene fragment was duplicated and it accounted for around 40% of TE-expanded genes. For example, a 9797 bp long maize CACTA element carried a 7923 bp gene GRMZM2G342226 (Fig. 2B), which encodes a UDPGP protein. In type 2, only 3′ -region of a gene was duplicated by a TE, accounting for 17.1% of TE-related genes (Fig. 2C). An example is the maize Helitron element, which carried a 3′ -region of the SuSy gene GRMZM2G139157. This 3′ -region encodes an intact SuSy domain. In contrast, in type 3, 5′ -region of a gene was duplicated by a TE and around 28.6% of TE-related genes were expanded by this type (Fig. 2D). For example, a strawberry LTR retrotransposon carried a 5′ -region of the gene gene00357-v1.0-hybrid, which encodes a SPP protein. In type 4, a middle region of a gene was duplicated by a TE and up to 14.3% of TE-related genes were expanded by this type (Fig. 2E). An example is the rice PACK-MULE on chromosome 1, which carried a fragment of a UDPGP gene with the locus name LOC_Os01g15910.
Different selection forces between monocot and dicot plants for the SPP family. Since the distinct difference was observed in their expansion history among the four gene families or in the same family between dicot and monocot plants (Figs 1 and 2), we investigated whether they were under different selection forces. We first identified reciprocal best matches for all four gene family members either from dicot or monocot plants by phylogenetic analysis using corresponding domain regions. The domain regions from these identified matches were then used to calculate their Ka (nonsynonymous substitutions per site), Ks (synonymous substitutions per site) and their ratios according to the description in Methods. We first analyzed and compared the Ka/Ks frequency distribution of these four gene families between monocot and dicot plants (Fig. 3A,B). For the SuSy, SPS and UDPGP families, similar frequency distributions of the Ka/Ks ratios were observed between monocot and dicot plants although their distribution patterns showed differences. There was also no significant difference between monocot and dicot plants in the average Ka, Ks and Ka/Ks values for these three gene families (Fig. 3C). For the SuSy family, most of the mass were centered near Ka/Ks > 0.5 whereas most of the mass for both SPS and UDPGP were centered near Ka/Ks = 0.1. Furthermore, the SuSy family also has the higher average Ka/Ks ratios at 0.87 for dicot and 1.04 for monocot plants whereas the average ratios for both SPS and UDPGP families were less than 0.2 for either dicot or monocot plants (Fig. 3C). These data suggested that the SuSy family evolved faster than the SPS and UDPGP families. However, for the SPP family, significant difference was observed between monocot and dicot plants in their Ka/Ks frequency distribution ( Fig. 3A,B). In dicot plants, most of the mass was centered near Ka/Ks = 0.2 whereas the peak was near 0.1 for monocot plants. The average Ka/Ks for dicot plants was 0.47, statistically higher than the ratio 0.14, which was calculated for monocot plants (Fig. 3C). To further analyze the reason why both dicot and monocot plants exhibited the difference in their evolutionary rate, we compared their Ka and Ks values separately. We found that the Ka value in monocots is twice the value in dicot plants whereas the Ks value in monocots is more than three times higher than that in dicots (Fig. 3C). As a result, dicots showed significantly higher Ka/Ks ratio than that in monocots. Thus, the SPP family in dicot plants evolved faster than that in monocot plants.
After analyzing the Ka/Ks ratios between the orthologous pairs, we further surveyed the selection forces in each amino acid site of corresponding domain regions of four gene families among 15 species. We submitted both aligned cDNA sequences deduced from domain amino acids and their phylogenetic tree to the sitewise likelihood-ratio (SLR) analysis 40 . For the SuSy family, Ka/Ks ratios in each site of the domain regions among tested 15 species ranged from 0.02 to 0.65 and no site showed the ratio higher than 1 (Fig. 3D). Similar results were observed for the SPS (  Rapid gene expansion and expression divergence. Our data showed that the SuSy, SPS, SPP and UDPGP families were evolutionarily conserved under purifying selection (Fig. 3). Our data also showed that gene expansion occurred in all the four gene families (Fig. 2). How did these expanded genes survive with limited functional divergence? We further investigated the expression profiling of these genes. We selected two gene families for these analyses. One of them is the SuSy family from the apple genome and another one is the UDPGP family from maize. The SuSy family underwent the rapid expansion during the divergence of M. domestica from F. vesca. In F. vesca, a total of 6 members were identified whereas up to 31 members were annotated in the apple genome ( Fig. 2A). Tandem duplication played a key role in the expansion. After duplication, very low divergence was detected. We have selected 10 closely related pairs for Ka/Ks evaluation and the analysis showed that the Ka/Ks ratios were low and some of them had no non-synonymous substitution (Fig. 4A). Expression analysis showed that all of the 31 genes were expressed in at least one of sixteen tissues (Fig. 4B). There were a total of 6 tandem duplicated arrays as shown by different colors in Fig. 4A. Individuals from each array showed expression differentiation in at least one of sixteen detected tissues. We then analyzed the expression divergence between the 10 closely related pairs. The data showed that these pairs also exhibited expression differentiation in at least one out of sixteen tissues by statistical analysis (Fig. 4B).
Another selected family for expression analysis is the UDPGP family from the maize genome. The UDPGP family also experienced the rapid expansion during the divergence of maize from sorghum. In sorghum, only 5 genes were identified to encode UDPGP whereas the maize genome encodes up to 17 UDPGP genes. Our investigation showed that both segmental duplication and transposition/retrotransposition played important roles in the family expansion (Fig. 2). A total of 7 genes (41.2%) have been involved in segmental duplication and both CACTA and LTR retrotransposons have contributed to the expansion of 5 genes (29.4%) ( Fig. 2). They were clustered into different groups (Fig. 4C). Microarray data consisting of 60 tissues and RNA_Seq data consisting of 15 tissues were employed for expression analysis. No probe is available or no RNA_seq signal was detected for the gene GRMZM6G729818. Expression evidence was presented for all the remaining 16 UDPGP genes (Fig. 4D). We also found that expression divergence was observed among segmentally duplicated genes and among CACTA/LTR-expanded genes (Fig. 4C,D). On the other hand, the phylogenetic analysis identified a total of 7 closely related pairs (Fig. 4C). Some of these pairs also showed Ka/Ks ratios as low as 0.01. However, all of these pairs exhibited significant expression divergence as shown by statistical analysis (Fig. 4D).

Genome variations and expression profiling between grain and sweet sorghum species.
BT × 623 and Keller belong to the same species S. bicolor. However, they exhibited obviously different phenotypes 41 . The former is grain sorghum and the latter is sweet sorghum with higher sucrose content in its stem. Genome sequences for these two cultivars were publicly available 42 . Our HMM searches identified the same numbers of SuSy, SPS, SPP and UDPGP families in these two cultivars. Genome-wide variation analysis identified one SuSy, two SPS, one SPP and two UDPGP genes showed 100% homology in their genome sequences with no single nucleotide polymorphism (SNP), no insertion/deletion (Indel) and no structural variation (SV) (Fig. 5A). The other 11 genes contained either SNP or Indel and only one gene Sobic.002G291200 has been detected with SV (Fig. 5A). The Ka/Ks analysis showed that up to 12 out of 18 genes from the four gene families showed no non-synonymous substitution with Ka/Ks = 0. Most of the remaining genes showed low ratios (0.043-0.562). Only one gene Sobic.006G213100 exhibited Ka/Ks > 1 (1.574) (Fig. 5A), suggesting the positive selection for this locus in these two varieties. As our data have showed that no positive selection occurred for all the four gene families among 15 species ( Fig. 3D-G), we were interested in whether the positive selection could be detected in other sorghum lines within the species. We have further confirmed that the gene Sobic.006G213100 showed no positive selection among 15 different species (Fig. 5B). We then analyzed the selection force of this gene among 45 intra-species lines, which belong to the same species S. bicolor. As these lines are from the same species, we could align the full-length amino acid sequences instead of domain sequences for the Ka/Ks analysis. Our data showed that a total of 7 sites were under positive selection with Ka/Ks ratios larger than 1(P < 0.05) (Fig. 5C). Three of them were located on domain region and the remaining 4 sites were out of the domain (Fig. 5C). The Ka/Ks ratios in the remaining sites were zero with no nonsynonymous substitution. The data suggested that mutation occurred only in limited sites and positive selection of the gene Sobic.006G213100 occurred during the divergence of intra-species.
Besides the genome divergence between BT × 623 and Keller, we also carried out the investigation of their expression profiling by quantitative reverse transcription PCR (qRT-PCR). We compared their expression abundance during seedling stage between BT × 623 and Keller (Fig. 5D-G). In each of analyzed four gene families, there was one gene which showed the most abundant expression level with at least 3 times higher than the remaining genes in that family. Further analysis showed that the highest expression level was higher than the sum of the remaining members. However, we did not observed any difference in their expression abundance at this growth stage between BT × 623 and Keller. We further analyzed the difference in expression regulation under sucrose treatment (Fig. 5H). Generally, for most of genes from these four gene families, no difference was observed between BT × 623 and Keller in their responses to sucrose stress. For example, in the SuSy family, no gene was detected with up-or down-regulation by sucrose treatment. In the SPS family, the gene Sobic.003G403300 was down-regulated by sucrose in both BT × 623 and Keller. In contrast, the SPP gene Sobic.009G041000 was up-regulated by  The phylogenetic tree of the SuSy family in apple. The apple family was expanded only by tandem duplication. Locus names with black fonts indicate that these genes were not duplicated. A total of 6 tandem clusters were identified and they were labelled with blue, brown, green, white with blue background, red and pink colors. (B) The expression profiling of this gene family among 16 tissues. Expression data were downloaded from the NCBI datasets (Methods) with accession number GSE42873. These issues were listed as below: 1, Flower_M67; 2, Fruit_M74_100 daa; 3, Flower_M74; 4, Fruit_M20_100 daa; 5, Leaf_M14; 6, Fruit_M20_harvest; 7, leaf_M49; 8, Fruit_M74_harvest; 9, Root_GD; 10, Stem_X8877; 11, Root_X8877; 12, Stem_GD; 13, Seedling_GD; 14, Seed_X4442 × X2596; 15, Seedling_ X4102; and 16, Seed_X3069 × X922. (C) The phylogenetic tree of the UDPGP family in maize. Locus names with red and green fonts indicate that they were from segmental duplication and TE-mediated expansion, respectively. The locus name with blue fonts indicates that the gene was expanded by TE but was located in a segmentally duplicated block. (D) The expression analysis of two datasets, which were downloaded from the NCBI datasets with accession numbers GSE27004 and SRA012297. These two datasets consisted of 60 and 15 tissues, respectively and their names were listed in their databases. In (A,B) both Ka values and Ka/Ks ratios for these nodes between closely related genes were calculated and were shown in the left and right of semicolon of the node. Normalized expression values were calculated by Agilent GeneSpring GX 11.5 software and were used for heat map analyses as shown in the right panel. Red, black, and green colors indicated the normalized expression values with > 0, = 0, and < 0, respectively, in the matrix. The stars "*" indicated the statistical difference in their expression level between these two genes.
Scientific RepoRts | 5:17583 | DOI: 10.1038/srep17583 sucrose in these two varieties. For the UDPGP family, two genes showed the difference in their responses to sucrose stress between these two varieties. One of them is Sobic.004013500, which was down-regulated by sucrose stress only in Keller. Another one is Sobic.007G075500, which was up-regulated by sucrose stress in BT × 623 but down-regulated in Keller (Fig. 5H). No visible phenotypic change by overexpression of sorghum SuSy genes in Arabidopsis. To understand the biological functions of genes encoding these four gene families, we overexpressed these genes under the maize ubiquitin promoter in Arabidopsis. For the SuSy family, all the coding regions of 5 genes were isolated from the sweet sorghum Keller by RT-PCR and subcloned into the binary vector pCambio1300 for Agrobacterium-mediated transformation. A total of 17 to 19 independent transgenic T1 plants were propagated from T0 generation (Fig. 6A) based on Southern blot hybridization (Fig. 6B). At least 3 independent T2 plants with single T-DNA insertion were selected for further investigation (Fig. 6A,B). Our data showed that no visible phenotypic difference was observed between wild type (WT) Arabidopsis and overexpression transgenic plants. Here we further characterized the sorghum gene Sobic.001G378300 by investigating its promoter activities and the effects of overexpressing this gene in Arabidopsis. We first analyzed its promoter activities in the promoter::GUS transgenic plants. Around 2 Kb of the promoter region upstream of the start codon of this gene was fused with the GUS reporter gene and was then delivered into the Arabidopsis genome. The GUS staining results showed that the activities of the promoter could be observed in the whole plants (Fig. 6C) including leaves (Fig. 6D), roots (Fig. 6E) and siliques (Fig. 6F). Further observation showed that the gene showed higher expression level in the phloem tissue of nodes and veins (Fig. 6C,D). We then characterized the transgenic Arabidopsis lines by overexpressing this gene. A total of 5 independent lines were subjected to expression analysis, which showed that the gene was highly expressed in these transgenic plants (Fig. 6G). However, these transgenic lines showed no statistical difference in their biomass yield (Fig. 6H). We also did not observe the difference in brix degree. We then subjected these lines into stress treatments. Under 1-7% sucrose stress, both transgenic plants and WT showed no obvious difference (Fig. 6I). Similarly, no difference was observed under both glucose and NaCl treatments (Fig. 6J,K). In general, no morphological phenotype was observed under normal or stress conditions after overexpression of the gene Sobic.001G378300.
Overexpression of one of sorghum SPS genes retards seed germination in Arabidopsis under normal and sugar stress conditions. In sorghum, we have identified a total of 5 SPS genes. In most of cases, these genes were expressed in multiple tissues (Fig. 7A). However, in each gene, the expression abundance varied among different tissues. Three genes including Sobic.003G403300, Sobic.004G068400 and Sobic.010G205100 showed the highest transcript level at young leaves and the gene Sobic.009G233200 exhibited the highest expression abundance in stem (Fig. 7A). To better understand their biological functions, coding regions from three genes were amplified from sweet sorghum by RT-PCR for overexpression under the maize ubiquitin promoter in Arabidopsis (Fig. 7B). By molecular characterization, we identified 13, 4, and 15 independent T2 lines from the genes Sobic.004G068400, Sobic.009G233200 and Sobic.010G205100, respectively, which showed high expression level with single copy of T-DNA insertion (Fig. 7B). These lines were then subjected into phenotypic characterization. For both genes Sobic.004G068400 and Sobic.010G205100, no obvious phenotypic variation was observed after overexpression in Arabidopsis when compared with WT plants. Here we focused on the gene Sobic.009G233200. Based on Southern blot hybridization, we have selected 4 independent lines with Single copy of T-DNA insertion (Fig. 7C). These lines showed high expression level for the overexpressed gene in Arabidopsis (Fig. 7D). We then subjected these lines into morphological characterization. Our data showed that overexpression of Sobic.009G233200 significantly retarded seed germination (Fig. 7E-G and Supplementary  Fig. S3 A-C). Under normal growth conditions, up to 93% of Arabidopsis WT seeds germinated after 2-day inoculation on ½ MS media whereas significantly lower percentages of seeds germinated for all identified 4 independent transgenic lines. Here we showed the results from two lines indicated by red and green cylinders, respectively in the Fig. 7E-G; Supplementary Figure S3 A-C. Their germination rates in these two lines were only 66% and 35%, respectively, under normal growth media (Supplementary Fig.  S3A). Under 1-9% of sucrose stress, germination for both WT and transgenic seeds were significantly retarded. However, transgenic seeds showed more sensitive response to the stress. For example, after 7-day inoculation on 7% sucrose containing MS media, up to 80% of WT seeds germinated whereas only less than 10% seeds germinated for the transgenic lines (Fig. 7E,H). Similarly, germination retardation was also observed for the transgenic seeds when they were germinated under both glucose and maltose stresses (Fig. 7 F and G; Supplementary Fig. S3 B and C). Although the transgenic Arabidopsis plants showed retarded germination under either normal or sugar stress conditions during early days, these lines exhibited no significant difference after more days of incubation on ½ MS media (Fig. 7I,J). From two weeks after germination, we further measured plant height, root length, flowering time and biomass yield during the whole developmental stages and found that no statistical difference was detected between WT and transgenic lines.
As seed germination is regulated by both gibberellic acid (GA) and abscisic acid (ABA), we further surveyed the expression regulation of Sobic.009G233200 under these two hormone treatments. Our data showed that the gene was upregualted by both GA3 and ABA treatments ( Supplementary Fig. S4 A and  B). We then examined the effects of GA3 and ABA on seed germination. The analysis showed that germination retardation was neither rescued by GA3 nor further postponed by ABA and no difference was observed between WT and transgenic plants (ubiquitin::Sobic.009G233200) under various concentrations of GA3 and ABA treatments after 9 days of germination ( Supplementary Fig. S4 C and D). Overexpression of one of sorghum SPP genes retards seed germination in Arabidopsis only under high salinity stress conditions. In the sorghum genome, a total of 3 SPP genes have been identified (Table 1). We carried out the comparative expression analysis between grain (BT × 623) and sweet sorghum (Keller) lines by RT-PCR (Fig. 8A,B). The gene Sobic.004G151800 showed the similar expression patterns between these two cultivars among tested tissues. This gene was constitutively expressed in all tested tissues with similar transcript abundance. In contrast, the gene Sobic.009G040900 was not expressed in a few of tested tissues in both BT × 623 and Keller. In BT × 623, the strong expression was detected in young and mature roots as well as stems; no signal was detected in mature seeds and young panicles; the remaining 5 tissues showed relatively less expression level. In Keller, the strongest expression was detected in both young and mature leaves followed by young roots and stems. No expression in young panicles and very faint signal in mature roots and mature panicles were detected. More interestingly, the gene Sobic.009G041000 was tissue-specific and exhibited the expression divergence between grain and sweet sorghum. In BT × 623, this gene was expressed only in young and mature roots; however, it was leaf-specific in keller (Fig. 8A,B). In addition, we further explored whether these three genes were regulated by some stresses in their expression. The gene Sobic.004G151800 showed no difference in its expression level under various stresses including drought (polyethylene glycol, PEG), high salinity (NaCl), cold (Ice-water), glucose and sucrose (Fig. 8C). The gene Sobic.009G040900 was down-regulated by 30% PEG treatment but was up-regulated by both glucose and sucrose. Its expression was not significantly regulated by both high salinity and cold stresses. Interestingly, the gene Sobic.009G041000 was up-regulated by all the tested 5 stresses.  Currently, no data has been reported on functional characterization of these SPP genes in sorghum. We have isolated two of these SPP genes from sweet sorghum and overexpressed them separately in the Arabidopsis genome. Among these 12 and 13 transgenic T1 lines overexpressing Sobic.004G151800 and Sobic.009G040900, respectively, we identified 4 independent lines for each gene that were with single T-DNA insertion (Fig. 8D). Expression analysis showed that the gene Sobic.004G151800 was highly expressed in all 4 independent lines (Fig. 8E). We also detected weak expression signal in WT as this gene shows very high homology to the Arabidopsis ortholog in their coding regions and primers used for RT-PCR may also bind to the ortholog for amplification. Similarly, another gene Sobic.009G040900 was also highly expressed in 4 independent transgenic lines after overexpression (Fig. 8E). All these 8 lines contained only single copy of T-DNA insertion as shown in Southern blot hybridization (Fig. 8D). All these independent 8 lines were then subjected into phenotypic characterization. We first grew these transgenic lines on ½ MS media together with WT plants and then characterized their phenotypes by measuring their plant height, root length and biomass yield. However, no statistical difference was detected between WT and transgenic plants (Fig. 8F). Since these two sorghum genes were regulated by various stresses (Fig. 8B), we then subjected these lines into high salinity stress. We also could not observe the difference ( Fig. 8G; Supplementary Fig. S5A). For example, root length in both WT and transgenic plants from overexpressing either Sobic.004G151800 (Supplementary Fig. S5A) or Sobic.009G040900 (Fig. 8G) showed no significant difference under different concentrations of salinity stress. Similarly, no difference was also observed when these transgenic lines were subjected into sucrose, glucose and maltose treatments ( Supplementary Fig. S6 A-C). In addition, both WT and transgenic lines also showed no difference in their germination rates ( Fig. 8H; Supplementary Fig. S5B). Under the NaCl stress less than 100 mM, no effect was observed on seed germination when compared with WT ( Fig. 8I; Supplementary Fig. S5C). However, under 150 mM NaCl-containing media, one transgenic line overexpressing Sobic.004G151800 showed statistically lower germination rate but no difference was observed for the remaining transgenic lines ( Supplementary Fig. S5C). Interestingly, under 150 mM NaCl-containing media, all transgenic lines over-expressing Sobic.009G040900 showed significantly lower germination rates (Fig. 8I). The data suggested that the gene Sobic.009G040900 might play a role in the seed germination under high concentration of salinity stress.
Similar to the SPS gene Sobic.009G233200, we surveyed the expression regulation of Sobic.009G040900 under GA3 and ABA treatments. The expression analysis showed that the gene was also upregulated by both GA3 and ABA treatments ( Supplementary Fig. S7 A and B). Furthermore, we examined the effects of GA3 and ABA on seed germination in WT and transgenic plants ubiquitin::Sobic.009G040900. The investigation showed that no difference was observed under various concentrations of GA3 and ABA treatments ( Supplementary Fig. S7 C and D).

Transgenic plants by overexpressing one of sorghum UDPGP genes in Arabidopsis increased biomass yield and flowered earlier.
The sorghum genome encodes 5 members of the UDPGP family ( Fig. 2A). We investigated their expression patterns by RT-PCR using total RNA samples from both BT × 623 and Keller (Fig. 9A). Except for the gene Sobic.002G291200, which was expressed in all tested six tissues including roots, stems, young and mature leaves, young and mature panicles, all the remaining 4 genes showed differential expression between BT × 623 and Keller in either tissues or transcript abundance. For example, the gene Sobic.004G013500 was mainly expressed in both mature leaves and young panicles in BT × 623 whereas its expression in Keller was mainly detected in both roots and young leaves. Another example is for the gene Sobic.010G251200, which was constitutively expressed in all 6 tissues with similar abundance in BT × 623. On contrast, in Keller, the gene was differentially expressed with obviously variable abundance in these 6 tissues. On the other hand, we also carried out the expression analysis of these genes under various stresses. We surveyed the expression regulation under drought (30% PEG), high salinity (250 mM NaCl), cold (ice-water), glucose (5%) and sucrose (5%). Under drought, high salinity, cold and glucose treatments, all analyzed four genes showed no significant difference in their expression level (Fig. 9B). However, under sucrose treatment, both genes Sobic.004G013500 and Sobic.007G075500 were down-regulated in their transcript abundance (Fig. 9B).
To further understand the biological functions of these 5 sorghum UDPGP genes, we overexpressed them in Arabidopsis. We have generated more than 20 independent T2 plants for each gene, from which we have selected at least 4 independent lines with single copy of T-DNA insertion and with high level of expression abundance (Fig. 9C). All of these lines were subjected into phenotypic characterization. However, our preliminary data showed that no obvious difference was observed in these transgenic lines when compared with WT plants in plant height, root length, flowering time and biomass yield for four genes including Sobic.002G291200, Sobic.004G013500, Sobic.007G075500 and Sobic.010G251200. We therefore focused on the remaining gene Sobic.006G213100. A total of 19 T2 plants were subjected into Southern blot hybridization for T-DNA copy number detection, from which we have identified 4 independent lines with single copy of T-DNA insertion (Fig. 9D). The gene Sobic.006G213100 showed high transcript abundance after overexpression in these four transgenic lines (Fig. 9E). These lines were grown on ½ MS media for phenotypic measurement, which showed that transgenic roots grew faster (Fig. 9F). The root length from all the 4 independent lines was statistical longer than that in WT plants (Supplementary Fig. S8A) Fig. S8B). However, at the mature stage (65-day-old), no significant difference was observed in their fresh weight between WT and transgenic plants (Supplementary Fig.  S8C). Interestingly, at the flowering stage, transgenic plants flowered earlier than WT plants (Fig. 9G). When the bolting rate of WT plants reached 81%, the rate in transgenic plants ranged from 87% to 94% (Fig. 9H). The data implied that the gene might play a role in flower timing. Discussion Evolutionary origins of SuSy, SPS, SPP and UDPGP gene families. In this study, we have genome-widely identified four gene families in 15 sequenced genomes. Besides these, we have also genome-widely identified these families in additional 35 species including 2 moss, 6 algae and 27 dicot/ monocot plant species, whose whole genome sequences are available (Supplementary Tables S1-S4). In addition, we also detected the distribution of these four gene families in nature by searching both Pfam and Interpro databases. Based on our searching results, the SuSy, SPS and SPP gene families generally presents in higher plants, mosses and bacteria, in which the organelle for photosynthesis has been evolved. All the three family members have been found in bacteria. However, their genomes usually encode only one member in each family and not all bacteria required these family proteins. All tested higher plants including mosses were found to encode at least one member of SuSy, SPS or SPP families. Thus, SuSy, SPS and SPP members are ubiquitous in higher plants. The SPP family also ubiquitously exists in algae. However, no SPS member was detected in all tested algae species. For the SuSy family, we genome-widely identified the family members in 6 sequenced algae genomes including Chlamydomonas reinhardtii, Coccomyxa subellipsoidea C-169, Micromonas pusilla CCMP1545, Micromonas sp. RCC299, Ostreococcus lucimarinus and Volvox carteri. However, only two species Chlamydomonas reinhardtii and Coccomyxa subellipsoidea C-169 encoded SuSy genes and no SuSy gene was detected in the remaining 4 genomes. On the other hand, the SPP family was also found in a few of fungus species and both SuSy and SPP members can be detected in Archaea. By searching the Pfam and Interpro databases, we obtained one sequence from marine thaumarchaeote (archaea) with accession number A0A075I6C7, which encodes an ornithine carbamoyltransferase. The sequence showed high homology to the SuSy domain. We have also detected two sequences from archaea encoding the SPP domain. The presence of the SPP family was also detected in some species of fungi. However, no sequence was detected in either archaea or fungi to encode a SPS protein (Supplementary Table S5). SPS proteins contain both domains with Pfam ID PF00534 and PF05116 (Methods). In green algae, both proteins with Pfam ID either PF00534 or PF05116 were detectable. However, no protein was identified to have these two domains. Thus, no SPS existence in green algae might be due to that no domain combination occurred in these two domains. On the contrary, the UDPGP family exists not only in monocot/dicot plants, mosses, algae and bacteria but also in animals and fungi although we could not detect any member in both archaea and virus (Table S5).
Previous data showed that archaea genomes did not encode any members of these sucrose synthesis related families 17 . However, we have detected homologous sequences in archaea to encode both SuSy and SPP domains (Supplementary Table S5). On the other hand, SuSy and SPP were widely detected in photosynthetic bacteria. Thus, based on our data, the SuSy and SPP families might be originated from archaea and then singular cell organism with the organelle for photosynthesis. Comparing to SuSy, SPS and SPP families, the UDPGP family exhibited distinct difference in its evolutionary history. It presents not only in plants but also in animals (Supplementary Table S5). Thus, our data showed that these four gene families underwent different origin and evolutionary histories.
For all four gene families, a genome from archaea, bacterium or fungus usually encodes only one member in each family and higher plant genomes encode various sizes of family members. At the early stage of evolutionary history, very low expansion occurred and a genome generally encodes a few members. This situation existed until the divergence between monocots and dicots (Fig. 1). A large scale of expansion occurred only for some species during the divergence from one species to another. As a result, different species encode different sizes of family members ranging from 4 to more than 30 genes ( Fig. 1 and Table 1). During the evolution, domain combinations were frequently observed among the domains from SuSy, SPS and SPP. Domain combinations were the processes that generated new genes and functional divergences 43 (Apic et al. 2001). One of the examples is the evolution of SPP enzymes, which consist of at least two domains. One is named as S6PP (PF05116) and another is Glycos_transf_1 (PF00534). The former comes from SPP enzymes and the latter comes from glycosyl transferases. Their combination led to the birth of SPS genes. Although both SuSy and SPP genes were detected in both algae and archaea, no SPS gene was evolved as no domain combination occurred between S6PP and Glycos_transf_1 domains in algae or archaea.

Expansion history and mechanisms of SuSy, SPS, SPP and UDPGP gene families. Although
four gene families exhibited the difference in their expansion patterns, they also showed some similarities. Before the divergence between monocot and dicot plants, four gene families experienced very low expansion and a MRCA genome encoded only one (for SPP), two (for SuSy and SPS) or three (for UDPGP) genes. After the divergence between dicots and monocots, monocot plant expanded faster than dicot plants as more members were detected in the MRCA among 4 monocot species than that among dicot species (Fig. 1). However, a large scale of expansion occurred during species divergence for both dicot and monocot plants as their MRCA family sizes were very small when compared with current family sizes ( Fig. 1 and Table 1). The data suggested the recently expansion and evolutionary events for these gene families. Furthermore, the expansion was also detected during the divergence of sub-species. One of the examples is the expansion of additional UDPGP member in indica rice 93-11. Generally, the japonica rice Nipponbare genome encodes a total of 6 UDPGP genes while a total of 7 genes are encoded in the indica genome. The additional member BGIOSGA003207 was recently tandem duplicated Scientific RepoRts | 5:17583 | DOI: 10.1038/srep17583 from BGIOSGA003204 after the divergence between indica and japonica. Different sub-species might experience different expansion patterns. As a result, some members might be expanded in a sub-species but not in another sub-species, which might contribute to intra-species divergence and the formation of sub-species. For example, the japonica rice genome encodes 7 SuSy genes including LOC_Os02g58480, LOC_Os03g22120, LOC_Os03g28330, LOC_Os04g17650, LOC_Os04g24430, LOC_Os06g09450 and LOC_Os07g42490. However, the indica rice genome encodes 8 members. Phylogenetic analysis showed that the MRCA between two sub-species contained 6 SuSy genes (Supplementary Fig. S9A). During the divergence of these two sub-species, the member located on chromosome 2 tandemly duplicated and gave birth to two members while only one member was detected in the japonica genome ( Supplementary  Fig. S9 A and B). A similar situation was also observed on chromosome 6 ( Supplementary Fig. S9 A and  C). However, on chromosome 4, one more gene was expanded, which occurred in the japonica genome ( Supplementary Fig. S9 A and D). As a result, the indica rice genome encodes 6 SuSy genes instead of 7 members in japonica.
Many mechanisms have been involved in gene expansion. Among them, genome-wide duplication significantly contributed to gene expansion 44 . In this study, we have detected the contribution of genome-wide duplication to the expansion of both SuSy and UDPGP gene families in soybean, apple and maize ( Fig. 2A,B). Besides the whole genome duplication, both tandem and segmental duplications also significantly contributed to gene expansion 45 . Our data showed that segmental duplication contributed to the expansion of four gene families in most of species. However, tandem duplication contributed to the gene expansion only in a few species. In addition to these, we also explored the contribution of transposons/retrotransposons to the expansion of these four gene families. We found that transposons/ retrotransposons played roles in the expansion of a few genes in some gene families. They might not contribute to the large scale of expansion of these four gene families. Generally, multiple molecular mechanisms have been involved in the expansion of these four gene families. Comprehensive expansion patterns have been observed, which varied among different gene families and different species. The data might imply that species-specific functions have been evolved in some gene families.
Expanded genes from various expansion mechanisms also frequently lost during long evolution due to redundant functions 45 . The fact might explain why both SPS and SPP gene families were not obviously expanded. After gene expansion, extra copies of genes might evolve into pseudogenes or genes with subfunctions/new functions. The Ka/Ks analysis showed that different gene families exhibited different rates of Ka/Ks ratios, suggesting the different selection force among different families/species (Fig. 3). The newly born genes might also keep the similar functions but survive with differential expression patterns compared with their parental genes. Our data showed that no same expression pattern was observed among members from the same family (Fig. 4). These data suggested the contribution of expression divergence to the survival of genes after expansion.
Recent positive selection occurred only during intra-species divergence in the sorghum UDPGP family. Our data showed that the divergence of genes from 4 gene families among distantly related species has been driven by purifying selection (Fig. 3). We have also surveyed the divergence of these genes within the species Sorgum bicolor, which indicated that most of these genes were under purifying selection with Ka/Ks less than 1 (Fig. 5). However, we detected one of UDPGP genes Sobic.006G213100, which was under positive selection with Ka/Ks = 1.574 (Fig. 5A). To further investigate the presence of the positive selection among a larger population within the species, we analyzed a total of 44 sorghum lines, in which the whole genome re-sequencing data are available 46 . We used BT × 623 as the reference genome and found 19 alleles with Ka/Ks > 1 whereas the remaining 25 alleles with Ka/Ks far less than 1. We have detected a total of 7 positively selected sites in the whole coding region (Fig. 5C). Among them, 3 sites were within the UDPGP domain region. The data suggested that the gene Sobic.006G213100 should have been suffered from positive selection during the divergence within the species, which, in turn, implied the role of this allele in divergence of sub-functions of this gene. Gene variants referring to adaptive phenotypes are positively selected 47 . Positively selected genes were frequently reported to play roles in signal transduction, cell-cell recognition, immune response, sexual reproduction, membrane and intracellular transporters 48 . For the UDPGP family genes, positive selection was also detected in the species Barbarea vulgaris 49 . In the species, two genes UGT73C10 and UGT73C11 were positively selected and they played roles in saponin-mediated insect resistance. Besides sorghum and Barbarea, we also investigated the presence or absence of positive selection of the UDPGP family genes in the Oryza sativa species as re-sequencing data from hundreds of thousands of lines from this species are available. We have analyzed a total of 1485 rice lines, whose re-sequencing data are available. However, no positive selection was detected in the whole coding region of a total of 6 UDPGP rice genes. Thus, positive selection in the UDPGP family occurred only during adaptive divergence within certain species. For these species, for example, both sorghum and Barbarea, positive selection should play important roles in adaptive divergence and evolution within species.
Newly revealed biological functions of SPS, SPP and UDPGP gene families. Sweet sorghum has been recognized widely as potential alternative source of bio-fuel due to its high fermentable sugar content in the stalk. However, little is known on sucrose metabolism related genes and their families in Scientific RepoRts | 5:17583 | DOI: 10.1038/srep17583 sorghum. In this study, we first genome-widely identified all genes encoding SuSy, SPS, SPP and UDPGP proteins and investigated their biological functions by overexpressing these genes in the Arabidopsis genome. For the sorghum SuSy gene family, we have generated transgenic Arabidopsis plants by overexpressing 5 SuSy genes (Fig. 6). However, no visible phenotypic variation was observed in these transgenic lines under normal or sugar stress conditions (Fig. 6). Bieniawska et al. (2007) analyzed biological functions of 6 Arabidopsis SuSy genes using corresponding T-DNA insertion mutants 25 . Their data showed that no obvious difference was observed between WT and mutants in growth phenotypes as well as in starch, sugar and cellulose content, seed weight or seed composition. Further investigation showed that sucrose synthase activity in the sus1/sus2/sus3/sus4 Arabidopsis mutant is sufficient to support normal cellulose and starch production 50 . Thus, knockout of one SuSy gene by T-DNA insertion could be complemented by other SuSy genes and as a result, exhibited no obvious phenotype. In addition, multiple studies were reported on overexpression of exogenous SuSy genes in the Arabidopsis or other genomes [18][19][20] . For example, overexpression of cotton SuSy gene in poplar increased cellulose content and crystallinity 19 . However, overexpression of the same gene in tobacco did not observe the similar phenotypes 18 . Thus, our and others' data suggested that plants might grow normally without visible change when one of their SuSy gene was knocked out or a SuSy gene was overexpressed.
Both SPS and SPP are key enzymes for sucrose synthesis. Overexpression of SPS or SPP significantly improved sucrose content and as a result, led to various phenotypes including biomass accumulation, promoting growth 27 , carbon metabolism 28 and improved fiber quality 26 . Reduced expression of SPS/SPP genes by RNAi or knockout by T-DNA insertion resulted in abnormal development 13,14,29 . In this study, we have overexpressed 3 SPS and 2 SPP sorghum genes in the Arabidopsis genome. We found that overexpression of a sorghum SPS gene Sobic.009G233200 in Arabidopsis retarded seed germination under normal and sugar stress conditions (Fig. 7). However, overexpression of a sorghum SPP gene Sobic.009G040900 in Arabidopsis postponed seed germination only under high salinity stress (Fig. 8).
No effect was observed for the remaining analysed SPS/SPP genes. To our knowledge, no data has been reported on the effect of SPS/SPP genes on seed germination. Thus, our data might provide some evidence for the new biological functions of SPS/SPP genes in seed germination.
Besides SuSy and SPS/SPP genes, we have also surveyed the biological functions of genes encoding UDPGP. We have generated transgenic Arabidopsis plants from 5 constructs by overexpressing 5 sorghum UDPGP genes (Fig. 9). Transgenic plants from 4 constructs showed no obvious difference when compared with WT plants. Only the transgenic plants by overexpressing Sobic.006G213100 showed higher biomass yield and flowered earlier than WT plants (Fig. 9). The increased biomass accumulation during vegetative growth stage was also observed in transgenic Arabidopsis plants by overexpressing an exogenous UDPGP gene from Larix gmelinii 37 . The earlier flowering by overexpressing an UDPGP gene was not yet reported. However, delayed flowering by silencing an UDPGP gene through RNAi has been observed 33 . In addition to the function in biomass accumulation and flowering, other functions of UDPGP genes have been reported including the roles in pollen callose deposition and male fertility 33,34 , which were not observed in this study.
In general, we have generated transgenic Arabidopsis plants by overexpressing 15 sorghum genes encoding SuSy, SPS, SPP or UDPGP enzymes. Most of the transgenic plants showed no obvious phenotype when compared with WT plants under either normal or sugar/salinity stress conditions. These might be partially due to that most of the enzymes catalyze reversible reactions, which might result in limited increase in sucrose synthesis. However, we still observed some phenotypes by overexpressing sorghum SPS, SPP and UDPGP genes in Arabidopsis. Both SPS and SPP genes showed the biological functions in seed germination, which were not reported previously and suggested the newly discovered function of sorghum SPS/SPP genes. The sorghum UDPGP gene Sobic.006G213100 was under positive selection and its overexpression in Arabidopsis exhibited its functions in biomass accumulation and shortening flowering time. The gene might play a role in adaptive evolution for changed ecological conditions during intra-species divergence. Thus, sucrose metabolism related genes play roles in not only sucrose synthesis and utilization but also being involved in other biological processes as well as adaptive evolution.

Methods
Plant materials and growth conditions. Both grain (BT × 623) and sweet (Keller) sorghum cultivars were collected from National Plant Germplasm System, United States Department of Agriculture (http://www.ars-grin.gov/npgs/). They were planted in greenhouse and were grown under natural light and temperature conditions. Abiotic stress and sugar treatments. The 14-day-old seedlings from both BT × 623 and Keller were collected from pots. After careful washing out soil from roots, the whole plants were subjected into sugar or various abiotic stress treatments. For sugar treatments, the whole seedlings were subjected to 5% glucose and sucrose solutions, respectively. Samples were then collected at 0, 2, and 8 h intervals for glucose treatment and 0, 2, and 6 h intervals for sucrose treatment for total RNA preparation. For drought, high salinity and cold stresses, seedlings were treated with 30% PEG, 250 mM NaCl solution or ice-containing water, respectively. Samples were collected at different time intervals (0, 0.5 and 2.0 h for drought and 0, 2 and 9 h for salinity and cold stresses) for total RNA isolation. Total RNA preparation and gene expression analysis. To survey the expression patterns among various developmental stages, tissues from different stages were collected including young leaf and root from 14-day-old seedlings, mature leaf, root and stem from 3-month plants, young and mature panicle from un-flowering and flowering plants, respectively. Samples from these tissues and various abiotic and sugar treatments were used for total RNA isolation using a QIAGEN (Hilden, Germany) RNeasy Mini Kit. The first-strand cDNAs were then synthesized from the total RNA samples using an Invitrogen kit. The synthesized cDNA samples were used for detecting expression patterns among different tissues or among various treatments. All polymerase chain reaction (PCR) reactions were carried out in 20 μ L reaction mixtures (20 ng cDNA, 200 μ M dNTPs, 2.5 mM MaCl 2 , 0.5 μ M primers and 1 unit QIAGEN Taq DNA polymerase and 1 × PCR buffer). The temperature profile for PCR is as follows: 94 °C for 2 min followed by 30 cycles at 94 °C for 10 s, 55-65 °C for 30 s (depending on annealing temperature of primer sets) and 7 °C for 1 min. The reaction was stopped by a 5 min extension step at 72 °C. PCR products were visualized by ethidium bromide staining in agarose gels. Photographs of the gels were recorded using Gel Documentation System from Bio-Rad. The qRT-PCR analysis was carried out using total RNA samples from both BT × 623 and Keller seedlings (two-week-old) according to our previous description 41 . Primer sequences used for qRT-PCR were listed in Supplementary Table S6.

Gene isolation, vector construction for promoter characterization and overexpression and
Agrobacterium-mediated transformation. Protein coding regions of 15 genes were amplified by RT-PCR with total RNA samples from the whole seedling plants (Keller) as templates. RT-PCRs were performed using QIAGEN One-Step RT-PCR Kit. The designed primer sets were then subjected to the phytozome sorghum database (http://phytozome.jgi.doe.gov/pz/portal.html#!info?alias = Org_Sbicolor) for a BLAST search to eliminate the non-specific primers. The Supplementary Table S6 lists all the primer sequences used for RT-PCR. PCR products were purified from agarose gels by QIAGEN QIAquick Gel Extraction Kit for sequencing. After sequencing verification, the amplified coding region fragments were subcloned into pCAMBIA1300 Ti-derived binary vector (CAMBIA, Canberra, Australia; http://www. cambia.org.au) under the control of the maize ubiquitin promoter. For promoter::GUS analysis, around 2 Kb of the sorghum Sobic.001G378300 promoter was amplified from the sorghum (Keller) genomic DNA using the primer set listed in the Supplementary Table S6. After verification by sequencing, the fragment was cloned in front of GUS reporter gene and then subcloned into pCAMBIA1300 Ti-derived binary vector. All the constructs were transformed into the Arabidopsis thaliana Columbia genome by the Agrobacterium-mediated method 51 .
For histochemical analysis, the GUS staining solution was composed of 0.02 M 5-bromo-4-chloro-3-indolyl-bb-D-glucuronide, 0.1 M NaH 2 PO4, 0.25 M ethylenediaminetetraacetic acid (EDTA), 5 mM potassium ferricyanide, 5 mM potassium ferrocyanide and 1.0% (v/v) Triton X-100 (pH 7.0). Various stages of leaves, flowers and roots were freshly collected from the promoter::GUS transgenic Arabidopsis plants (T1 and T2 generations) for the detection of GUS activities. These samples were incubated with the staining solution for overnight at 37 °C. Samples were then subjected to GUS activity observation and photographing after being decolorized by a series concentrations of alcohol.

Molecular and phenotypic characterization of transgenic plants. The copy numbers of T-DNA
insertion in transgenic Arabidopsis plants were analyzed by Southern blot hybridization. A total of 6 μ g DNA samples were digested with restriction enzymes PstI or SpeI and were then separated by 0.7% agarose gels. The separated DNA samples were transferred onto nylon membranes for probe hybridization. Probes were synthesized by PCRs with the DIG Probe Synthesis Kit (Roche), using the primers listed in Supplementary Table S6. Southern blot hybridizations were carried out according to the manufacturer's instructions (Boehringer Mannheim). Chemiluminescence was detected using CDP Star substrate (Roche Applied Science).
The transgenic lines with single copy of T-DNA insertion were subjected to expression analysis by RT-PCR using the gene-specific primer sets listed in Supplementary Table S6. Only these lines with single T-DNA insertion and with high expression level for the targeted gene were subjected to phenotypic investigation. Both WT and transgenic lines were germinated and grown under MS media with or without abiotic /sugar stresses. For biomass analysis, both WT and transgenic lines were also grown in soil pots under similar growth conditions. We have surveyed various phenotypes including germination rate, root length, plant height, biomass weight and flowering time according to the description.
Publicly available expression data and their processes. To investigate the expression divergence of SuSy, SPS, SPP and UDPGP genes in apple (M. domestica) and maize (Z. mays), some of microarray and RNA_Seq datasets were downloaded from the NCBI database (http://www.ncbi.nlm.nih.gov). For the apple Borkh, dataset with accession number GSE42873 were used, which represented transcriptional profiling of 32 different tissues. For maize B73, microarray data with accession number GSE27004 was downloaded, which consisted of 180 samples. As no microarray probe is available for some of maize SuSy, SPS, SPP and UDPGP genes, we also downloaded RNA_seq data with accession number SRA012297 for expression analysis of these genes with no microarray probe. After downloaded, both microarray and RNA_Seq data were processed and analyzed by using the CLC Genomics Workbench software (CLC Bio, Swansea, UK). Expression divergence between most recently expanded genes was determined by Student's t-test. These genes with a statistical difference at P < 0.05 were regarded as differentially expressed genes.
Genome sequences and their annotation as well as re-sequencing data. A total of 50 whole genome sequences and their annotation data were downloaded from the Phytozome database (Phytozome 10.1, http://phytozome.jgi.doe.gov/pz/portal.html). A total of 44 sorghum lines were re-sequenced 46 and their re-sequencing data were downloaded from the NCBI Short Read Archive database under accession numbers SRS378430 to SRS378473. For rice, the genome sequence data from a total of 1483 rice germplasm lines were available for download from the RiceVarMap database 52 (http://ricevarmap.ncpgr.cn/) and we used the data for our analysis.
Genome-wide identification of genes encoding SuSy, SPS, SPP and UDPGP. All the SuSy and UDPGP proteins contain a conserved domain with Pfam (http://pfam.sanger.ac.uk/) ID PF00862 and PF01704, respectively. All SPS proteins contain both domains with Pfam ID PF00534 and PF05116. However, SPP proteins have only one of them with Pfam ID PF05116. Thus, we identified these four gene families based on their domain constitute and sequences. We have employed both profile HMM and BLAST searches to genome-widely identify all putative genes encoding SuSy, SPS, SPP and UDPGP enzymes in all the 50 genomes. To build HMM profiles, representative domain protein sequences for PF00862, PF01704, PF05116 and PF00534 were retrieved from the Pfam database. These domain sequences were then aligned using ClustalX 2.0 (www.clustal.org) and were then used to generate HMM profiles for HMM searches with E-value cutoff of 1.0 against annotated protein databases from all the 50 genomes. Besides the HMM searches, BLASTP searches were also performed using the domain sequences as queries to figure out more members in corresponding genomes. These identified members from both HMM and BLASTP searches were then submitted to domain searches to confirm the presence of corresponding domains with E-value = 0.01 as cutoff level.
Domain amino acid sequence alignment and phylogenetic analysis. The program ClustalX 2.0 was used for the domain amino acid sequence alignment, which was used for the construction of phylogenetic trees according to the description by Jiang et al. (2008) 53 . For the SuSy, SPP and UDPGP families, amino acid sequences from only one corresponding domain were used for alignment and tree construction. For the SPS family, both domains PF00534 and PF05116 were used for sequence alignment and phylogenetic analysis.
Detection of tandemly and segmentally duplicated genes in 15 genomes. Tandemly duplicated genes were determined by sequence identity and physical positions of corresponding genes. They should meet the following criteria: (1) aligned regions should cover at least 30% of query sequences by BLAST search using BLOSUM62 matrix with an E value of less than 0.01; (2) sequence identity should be 70% or greater; (3) the duplicated genes should be less than or equal to 10 genes apart; and (4) tandem pairs should be within 100 Kb for algae and Arabidopsis and 350 Kb for the remaining genomes as suggested by Lehti- Shiu et al. (2009) 54 .
Segmentally duplicated genes were determined if these genes were located on segmentally duplicated chromosome blocks. Sequence regions franking 50 kb upstream and downstream of targeted genes were submitted to the identification of segmentally duplicated blocks by using the DAGchainer program 55 . A minimum of 5 gene pairs was used to define a block.

Detection of transposon/retrotransposon-related expansion of SuSy, SPS, SPP and UDPGP genes.
To survey the contribution of retrotransposons and DNA transposons to the expansion of the SuSy, SPS, SPP and UDPGP gene families, we identified all these mobile elements in the flanking genomic sequence regions of the 50 kb upstream and downstream of targeted genes. We only investigated the contribution of full-length retrotransposons to the gene family expansion as it is difficult to determine the contribution of non-intact retrotransposons. We identified the full-length retrotransposons by the LTR_ FINDER program 56 . If a gene was located within a full-length retrotransposon the gene was regarded as retrotransposon related gene. Besides retrotransposons, we also surveyed the contribution of retrogenes to the family expansion. To detect possible retrogenes in these four gene families, genes with only single exon were selected and their amino acid sequences were subjected to BLASTP searches against all of the remaining proteins encoded by genes with two or more exons. Base on the BLASTP results, retrogenes were identified according to the criteria as described by Wang et al. (2006) 57 .
To determine the contribution of DNA transposons to the gene family expansion, the same flanking genomic sequences of targeted genes were submitted to the identification of four major transposon family members, which were mutator-like transposable element (MULE), hAT, CACTA and Helitron families. We identified these elements according to the previous description 58 . Pairs were primarily selected according to phylogenetic trees. Only alignments longer than 150 bp with at least 70% identity were selected when incomplete overlap were found in a pair. Both Ka and Ks values were then estimated from their aligned cDNA sequences using the yn00 program of the PAML4b package 59 .

Estimation of
To detect purifying/positively selected amino acid sites, phylogenetic trees were first constructed. The amino acid alignment, which was used for tree construction, was then transferred to the original cDNA sequences. Both phylogenetic trees and cDNA sequence alignment were used to detect corresponding amino acid sites with purifying/positive selection according to the "sitewise likelihood-ratio" (SLR) method 40 .
To determine positive selection within intra-species, re-sequence data from 44 sorghum lines 46 and 1483 rice germplasm lines 52 were used to achieve cDNA sequences and their corresponding protein sequences for all the four gene family members. Based on these sequence data, the positively selected sites were detected by the SLR program.